Fix #414: Crash when re-applying time signature to parts (out of bounds)

[worldwideweary]

Resolves: #414

Test it out, seems to work okay

[worldwideweary]

Obviously I'm not spot on because there's

27 - tst_measure (Failed)

... don't have local testing here though so not sure what that is yet

[worldwideweary]

What's the deal with the failing mtest?

I dunno man. I don't know much about the unit testing... so I'm curious myself. I don't see how something that is out of bounds could be "appropriate" as if that would alter some results, so maybe it's having to do with using the "score" instead of this pointer

[worldwideweary]

There's the defined constructor which this code calls (takes a score pointer), and the old one which used the "default" constructor which doesn't have any code here (but can of course take a pointer to its own type) ... the one this uses and the only one showing in the codebase does:

TimeSig::TimeSig(Score* s)
  : Element(s, ElementFlag::ON_STAFF | ElementFlag::MOVABLE)
      {
      initElementStyle(&timesigStyle);

      _showCourtesySig = true;
      _stretch.set(1, 1);
      _sig.set(0, 1);               // initialize to invalid
      _timeSigType      = TimeSigType::NORMAL;
      _largeParentheses = false;

So maybe these default attributes are not in accord with the default construction in some way that alters the mtest results... like maybe the styles are diferent from initElementStyle, or showcourtesy sig is false or something

[Jojo-Schmitz]

  _sig.set(0, 1);               // initialize to invalid

Esp. this, leads to:

--- /home/runner/work/MuseScore/MuseScore/mtest/libmscore/measure/measure-6-ref.mscx    2024-09-12 10:58:03.497486946 +0000
+++ measure-6.mscx  2024-09-12 11:19:15.121886888 +0000
@@ -59,8 +59,10 @@
       <Measure>
         <voice>
           <TimeSig>
-            <sigN>4</sigN>
-            <sigD>4</sigD>
+            <sigN>0</sigN>
+            <sigD>1</sigD>
+            <stretchN>0</stretchN>
+            <stretchD>1</stretchD>
             </TimeSig>
           <Chord>
             <durationType>quarter</durationType>

[Jojo-Schmitz]

Here's what that mtest does:

void TestMeasure::spanner_A()
      {
      MasterScore* score = readScore(DIR + "measure-6.mscx");

      score->select(score->firstMeasure());
      score->startCmd();
      score->localTimeDelete();
      score->endCmd();
      score->doLayout();
      QVERIFY(saveCompareScore(score, "measure-6.mscx", DIR + "measure-6-ref.mscx"));
      delete score;
      }

I don't claim to understand this test... Esp. how this test differs from 5, 6, 7, 8 and 9

[worldwideweary]

You know what, I think I see what the problem is already... The results of the dude's score gives:

0/1

Something's wrong elsewhere... but it still doesn't make sense yet

[worldwideweary]

Oh, maybe it's the stretch: nts->setStretch(nts->sig() / mAfterSel->timesig());

nts->sig() is whatever was defaulted. I'll set it to the lastDeletedForThisStaff and see how the tests go

[worldwideweary]

OK, hopefully that takes care of it. If there are any problems in the future having to do with this, alternatively the code could just use the "this" pointer like before but then specifically setScore(score) afterwards.